can wages buy honesty? the relationship between relative ... wages buy honesty.pdf · ternative...

DOI: 10.1111/j.1475-679X.2012.00456.xJournal of Accounting ResearchVol. 50 No. 4 September 2012

Printed in U.S.A.

Can Wages Buy Honesty? TheRelationship Between Relative

Wages and Employee Theft

C L A R A X I A O L I N G C H E N ∗ A N D T A T I A N A S A N D I N O †

Received 1 September 2011; accepted 25 December 2011

ABSTRACT

In this study, we examine whether, for a sample of retail chains, high levels ofemployee compensation can deter employee theft, an increasingly commontype of fraudulent behavior. Specifically, we examine the extent to which rel-ative wages (i.e., employee wages relative to the wages paid to comparableemployees in competing stores) affect employee theft as measured by inven-tory shrinkage and cash shortage. Using two store-level data sets from theconvenience store industry, we find that relative wages are negatively associ-ated with employee theft after we control for each store’s employee charac-teristics, monitoring environment, and socio-economic environment. More-over, we find that relatively higher wages also promote social norms such that

∗Department of Accountancy, College of Business, University of Illinois at Urbana-Champaign; †Leventhal School of Accounting, University of Southern California. Ac-cepted by Phil Berger. We thank the editor, an anonymous referee, Paul Adler,Dennis Campbell, Willie Choi, James Detert, Bart Dierynck, David Erkens, Curtis Hall, AyseImrohoroglu, Steve Kachelmeier, Robert S. Kaplan, Simon Jonghwan Kim, F. Asis Martinez-Jerez, Ella Mae Matsumura, Ken Merchant, Kevin Murphy, Lamar Pierce, Mina Pizzini, KarenSedatole, Lloyd Tanlu, Bill Tayler, Kristy Towry, Wim Van der Stede, Greg Waymire, SallyWidener, Michael Williamson, Alicia Yancy, Mark Young, Christopher Yust, and seminar par-ticipants at Claremont McKenna College, Emory University, Erasmus University, Harvard Busi-ness School, Southern Methodist University, University of Arizona, and University of Texas atAustin, as well as participants at the 2010 Management Accounting Section Meeting, the 2010Annual Meeting of the American Accounting Association, and the 2011 International Sym-posium on Management Accounting for their helpful comments. We also thank the NationalAssociation of Convenience Stores for access to data. All errors remain our own.

967

Copyright C©, University of Chicago on behalf of the Accounting Research Center, 2012

968 C. X. CHEN AND T. SANDINO

coworkers are less (more) likely to collude to steal inventory from their com-pany when relative wages are higher (lower). Our research contributes to anemerging literature in management control that explores the effect of effi-ciency wages on employee behavior and social norms.

1. Introduction

Efficiency wage theories propose that higher levels of pay induce higherproductivity by motivating employees to exert greater efforts (out ofreciprocity or a desire to maintain high paying jobs) and/or by at-tracting higher quality employees (Malcomson [1981], Yellen [1984],Akerlof [1984], Shapiro and Stiglitz [1984], Akerlof and Yellen [1990],Fehr and Gachter [2000]). Prior studies have tested the association be-tween higher wages and employee effort and turnover.1 In this study, wecomplement the efficiency wages and management control literatures byexamining the impact of relative wages on employee theft, where rela-tive wages are estimated by comparing employee wages with the wages ofcomparable employees working for other similar organizations in the re-gion.2 Support for an association between relative wages and employeetheft would suggest that employee compensation levels can provide an al-ternative mechanism to deter fraudulent behavior, beyond other honesty-inducing control mechanisms studied in the accounting and control liter-ature (e.g., Hansen [1997], Chow, Cooper, and Waller [1988], Evans et al.[2001], Webb [2002], Zhang [2008], Hesford and Parks [2010]).

To the best of our knowledge, the only other empirical study that ex-plores the relation between compensation and employee theft using fielddata is Greenberg [1990], who finds that, for groups of employees whoreceived a 15% pay-cut, the level of theft (measured by shrinkage rates)increased in the post-pay-cut period, and was higher than that for groupswho did not receive the pay cut. However, there is a substantial differencebetween reducing employees’ pay levels and setting ongoing pay levels. In-deed, recent field studies suggest that the effect of pay cuts on employeeperformance is short-lived (Lee and Rupp [2007]). Our study is the first toexamine cross-sectional associations between employee theft and relativewages, where relative wages are estimated based on the local labor markets.

Employee theft is a major management control problem that resultsin up to $200 billion in losses in U.S. businesses every year (Murphy

1 For example, a series of laboratory experiments shows that employees exert higher effortwhen their employers choose to pay them higher wages than the employees expected (e.g.,Fehr, Kirchsteiger, and Riedl [1993], Fehr and Falk [1999], Hannan, Kagel, and Moser [2002],and Hannan [2005]). A handful of field studies have also shown that relatively higher wagesare associated with less shirking in the workplace, greater pay satisfaction, and lower likelihoodto quit (Cappelli and Chauvin [1991] and Levine [1993]).

2 Following Hollinger and Clark [1983, p. 2], we refer to employee theft as any “unautho-rized taking, control, or transfer of money and/or property of the formal work organizationthat is perpetrated by an employee during the course of occupational activity.”

CAN WAGES BUY HONESTY? 969

[1993]). This problem is particularly severe in retail chains, where geo-graphic dispersion leads to significant monitoring problems (Brickley andDark [1987], Campbell, Datar, and Sandino [2009]). A survey conductedby Hollinger and Clark [1983] indicates that 35% of U.S. retail store em-ployees admit to stealing from their companies, and, according to the 2008National Retail Security Survey, inventory-related employee theft amountedto a loss of $15.9 billion in the retail industry alone (NRSS [2008]).

Drawing on efficiency wage theories, we predict that relatively higherwages will discourage employee theft for two reasons: first, employees re-ceiving relatively higher wages are less inclined to commit theft as theyattempt to reciprocate positively to their employers and/or to retain theirhigh-paying jobs, while employees receiving relatively lower wages are moreinclined to commit theft due to a desire to retaliate against their employ-ers for treating them unfairly and/or disregard for their jobs (motivationmechanism); second, firms that offer relatively higher wages may attract ahigher proportion of honest workers (selection mechanism).

In addition to examining the overall effect of relative wages on employeebehavior, we examine whether relative wages affect social norms amonggroups of employees (in stores that are staffed by more than one employee)that could mitigate or instigate theft. We predict an interaction effect be-tween relative wages and coworker presence, such that coworkers are morelikely to monitor each other and reduce theft under relatively higher wages,but they are more likely to collude against their employer and increase theftunder relatively lower wages.

We test our predictions using two proprietary data sets from retail chainsand supplement our analysis with insights from telephone interviews withstore managers of nine convenience store chains in our sample. The firstdata set contains cross-sectional data on cash shortage and inventory shrink-age as a percentage of sales from the 76 stores of a mid-sized conveniencestore chain (hereafter, CS Chain) for the year 1999. The second data setincludes data on the ratio of cash shortage to sales from a subsample of 251stores (327 store-years) in 31 chains that completed a survey conductedby the National Association of Convenience Stores (hereafter, NACS) inthe years 2003 and 2004. Using both the CS Chain data set and the NACSdata set allows us to assess the generalizability of our findings across differ-ent samples. It also allows us to triangulate our tests against two differentmeasures of employee theft: cash shortage and inventory shrinkage, whichare the two most important sources of financial loss in the retail indus-try, accounting for losses of 0.25% and 1.60% of retail sales in the UnitedStates, respectively (NRSS [2005]).3 To control for the portion of inven-tory shrinkage and cash shortage that is due to incompetence or error of

3 Both measures have strengths and weaknesses. Inventory shrinkage accounts for a higherpercentage of losses and, according to Dwyer [1992], 70% of inventory shrinkage can be at-tributed to employee theft because the small size of this industry’s stores results in low inci-dence of customer shoplifting. Although cash shortage accounts for a lower percentage of


employees, we include employee skills and employee experience in our em-pirical tests.

We use store-level multivariate analyses to examine the relation betweenemployee theft and the explanatory variables of interest, namely, relativewages and the interaction between relative wages and coworker presence.We calculate relative wages by estimating the difference between employeewages at a given store and median employee wages for cashiers for salesorganizations in the Metropolitan Statistical Area (MSA) in which the storeis located. We estimate coworker presence by dividing the total annual laborhours reported for a store by the total annual opening hours of the store.

As predicted, we find a negative association between relative wages andemployee theft in both the CS Chain and NACS samples, after controllingfor each store’s socio-economic environment, monitoring environment,and employee characteristics. A cost-benefit analysis reveals that the ben-efits from reducing theft alone cover 39% of the costs associated with wageincreases.4 In subsequent analyses, we show that employee theft decreasesin the magnitude of overpayment but does not increase in the magnitudeof underpayment. This finding is contrary to the argument in prior litera-ture that underpaid workers retaliate against their employers in proportionto their underpayment but overpaid workers rationalize the overpaymentaway (Akerlof and Yellen [1990]).

We also find our predicted interaction effect between relative wages andcoworker presence. Specifically, we find that while coworker presence perse is associated with higher levels of inventory shrinkage, this association isreduced when relative wages are higher. Our results suggest that relativelyhigher wages mitigate worker collusion to steal inventory from the com-pany. We do not find this interaction effect, however, when employee theftis measured with cash shortage.

Our research contributes to the management control literature that ex-amines the effects of compensation premiums on employee behavior (e.g.,Hannan, Kagel, and Moser [2002], Hannan [2005], Kuang and Moser[2009], Matuszewski [2010]) by showing that relatively higher wages notonly affect employee effort but also discourage employee theft. Moreover,our results suggest that relatively higher wages not only have the direct ef-fect of curbing employee theft, but also promote an ethical environmentamong coworkers. In doing so, we also contribute to the literature on socialnorms and the influence of coworkers on the behavior of employees (e.g.,Jaworski and Young [1992], Robinson and O’Leary-Kelly [1998], Towry[2003], Zhang [2008], Hannan, Towry, and Zhang [2011], Tayler andBloomfield [2011]). Consistent with the insights from Tayler and Bloom-field’s [2011] experimental study, we provide empirical evidence from the

losses, it is arguably a cleaner measure of employee theft than inventory shrinkage because itis predominantly attributable to the workers’ actions rather than to customer shoplifting.

4 This result does not consider any other potential benefits from paying high relative wages,such as lower turnover and training costs, decreased shirking, and increased sales.


field that compensation practices can shape group norms, which, in turn,influence employee behavior.

Our study has important practical implications. It sheds light on anunderexplored mechanism through which a company can deter theft.The Committee of Sponsoring Organizations of the Treadway Commission(COSO) framework, which the majority (82%) of U.S. CFOs use as theframework for internal control in their companies (Shaw [2006]), empha-sizes that adequate human resource practices such as “competitive com-pensation programs” (COSO [1994, p. 29]) play an important role inpreventing fraud. Our results provide support for this guideline and of-fer useful insights to management accountants and CPAs, who are play-ing an increasingly active role in the prevention and detection of inter-nal fraud and theft in firms (Wells [2001, 2002]). In addition, AuditingStandards AU Section 316 (Auditors’ Consideration of Fraud) suggeststhat auditors should consider the incentives, opportunities, and atti-tudes/rationalizations relevant to asset misappropriation. Our findings pro-vide insights to auditors by identifying relative wages as a potential an-tecedent of both the incentives and attitudes/rationalizations relevant toasset misappropriation.

The remainder of this paper is organized into four sections. In section2, we develop our hypotheses. Section 3 discusses our data and researchdesign. Section 4 presents our empirical analyses and results. Section 5concludes.

2. Hypotheses Development

2.1 RELATIVE WAGES AND EMPLOYEE THEFT

Drawing on the efficiency wages literature, we predict that relativelyhigher wages should reduce employee theft for three reasons5:

(1) Relatively higher wages induce employees to reciprocate positively totheir employers, making it less likely that employees will commit theft.

(2) Relatively higher wages increase employees’ costs of being fired, in-creasing the cost of theft.

(3) Relatively higher wages attract more honest employees.

The first argument, which suggests that relatively higher wages leadto lower theft due to reciprocity considerations, relies on insights fromefficiency wage theories based on sociological models, such as the “gift

5 The efficiency wages literature assumes that different firms may offer different wages toemployees with identical observable characteristics for a number of reasons (e.g., Brown andMedoff [1989], Cappelli and Cascio [1991]). For example, Yellen [1984, p. 201] argues that“if the relationship between wages and effort differs among firms, each firm’s efficiency wagewill differ, and, in equilibrium, there will emerge a distribution of wage offers for workers ofidentical observable characteristics.”


exchange” and “reciprocity” models. These theories suggest that employ-ees judge the fairness of their employment by comparing their wages withthose of comparable employees (Adams [1963], Akerlof [1982, 1984], Cap-pelli and Chauvin [1991]). According to these theories, employees whobelieve they are overpaid are likely to reciprocate to their employers byworking harder and developing a productive environment; in contrast, em-ployees who believe they are underpaid are likely to feel entitled to shirk or,according to this study, commit theft, to retaliate against their employers.

The second argument is based on efficiency wage theories such as the“shirking model,” which postulate that above-market wages deter employ-ees from shirking due to the high costs of losing their jobs. It is arguedthat employees exert effort and avoid improper behavior so as to retaintheir jobs and continue earning above-market rents (Shapiro and Stiglitz[1984], Baker, Murphy, and Jensen [1988]). According to these theories,firms might benefit from paying above-market wages due to increases inproductivity and decreases in employee turnover and related turnover costs(Stiglitz [1974], Yellen [1984], Krueger and Summers [1988], Campbell III[1993]).6

While the first two arguments are due to motivation reasons, the third ar-gument is based on the selection model of efficiency wage theories, whichposits that, when workers are heterogeneous in their abilities, firms haveimperfect information about worker abilities, and there is a positive correla-tion between ability and workers’ reservation wages, firms that offer higherwages attract a better pool of applicants with higher ability (Weiss [1980],Malcomson [1981], Yellen [1984]). Even though the previous literature hasbuilt the selection argument primarily on the abilities of workers, the sameargument could apply to the integrity of workers, implying that firms thatoffer higher wages can attract higher quality employees who behave moreethically. It is reasonable to expect a positive correlation between honestyand workers’ reservation wages since dishonest workers are likely to havea lower reservation wage than otherwise comparable workers as they countnot only on the wage offered by the employer but also on the cash and/ormerchandise they could steal from the workplace if they accepted the job.The selection argument would not hold if: (1) employees stealing were con-sistently caught and terminated, or (2) firms were able to design screeningdevices to induce workers to reveal their honesty/integrity. However, it is of-ten hard for firms to detect employee theft, and, even though firms usuallyconduct background checks or honesty/integrity tests before they hire em-ployees (Bernardin and Cooke [1993]), these tests are unlikely to provideaccurate information about employees’ honesty.

6 If paying above-market wages were a profitable solution for all companies, then all com-panies would choose to pay such wages. However, employees’ incentives not to shirk wouldcontinue to apply because the higher wages would result in unemployment. Thus, the em-ployees’ reservation wage is lower than the employment wage (Shapiro and Stiglitz [1984]).


The theoretical arguments presented above are in line with insights fromfield studies and practitioner articles suggesting a negative relationshipbetween relatively higher wages and employee theft. Field studies con-ducted by Bewley [1999], Hollinger and Clark [1983], Nagin et al. [2002],and Victor, Trevino and Shapiro [1993] find that employees behave oppor-tunistically when they feel that their jobs are not worth keeping and/orwhen they have negative perceptions about the way their employers treatthem. For example, one business owner interviewed by Bewley [1999,p. 48] indicated: “I care about morale because, in the case of my secretary,she could embezzle money easily.” Since employees’ perceptions are influ-enced by relative wages (e.g., Levine [1993], Pfeffer and Langton [1993]),we expect relatively higher wages to be negatively related to employee theft.A field study that is closer to ours, conducted by Greenberg in 1990 andfollowed up by two experimental studies (Greenberg [1993, 2002]), showsa positive association between wage cuts (underpayment) and employeetheft.7 The practitioners’ literature also offers examples that suggest a neg-ative association between the level of employee pay and employee theft.For example, a BDO executive interviewed by the Wall Street Journal in-dicated that decreases in employees’ benefits lead them to feel unfairlytreated and to justify stealing from their companies (Needleman [2008]).Similarly, companies such as Costco, Trader Joe’s, and The Container Storeoffer wage premiums to their employees in order to deter theft (Speizer[2004], Greenhouse [2005]).

Although the theoretical arguments provided above suggest a negativerelationship between relative wages and employee theft, this relationshipmay not be observed in practice for two reasons. First, absent a salient eventsuch as a pay cut (as studied by Greenberg), employees may not activelybenchmark their wages against those of comparable employees. Second,prior studies show that, over time, employees may overlook wage deviationsas they reassess the value of their own inputs. For example, using airline on-time performance as a proxy for pilot effort, Lee and Rupp [2007] examinethe impact of a pay cut on airplane pilots’ effort and find that the negativeeffect of the pay cut on employee effort is short-lived and does not persistafter one week. Similarly, using field experiments, Gneezy and List [2006]find that workers are more productive in the first few hours in a job wherethey are led to believe that they are overpaid, but this productivity declinesto the level of non-overpaid employees after a few hours. Thus, whether ornot a relationship exists between ongoing relative wages and employee theftremains an empirical question. Based on the above discussion, we posit thathigher relative wages decrease employee theft. More formally:

7 Wage cuts in this case are measured relative to the employee’s own prior wage (in hisfield study Greenberg studies the effects of pay cuts of employees in a manufacturing firm) orrelative to the wage the workers were led to believe they would receive (in his experiments,Greenberg studies the effects of reducing wages relative to the wage that the experimentersinitially suggested to the participants).


H1: Relative wages are negatively associated with employee theft.

2.2 EFFECTS OF RELATIVE WAGES ON EMPLOYEES’ COLLECTIVE BEHAVIOR

Our second hypothesis examines the effects of relative wages on socialnorms and employees’ group behavior. Research in social psychology aswell as economics suggests that individuals are influenced by social norms.8

Social norms play an important role in individuals’ behavior because fail-ure to conform to social norms can lead to informal punishment, suchas exclusion from the peer group. For example, Robinson and O’Leary-Kelly’s [1998] field study of 187 employees from 35 groups in 20 orga-nizations documents a positive association between the antisocial behav-ior of employees and their coworkers’ antisocial behavior. In our researchsetting, because it is easier for coworkers to detect employee theft thanfor managers, social norms can be an important determinant of employeetheft.9

The main effect of coworker presence on employee theft is ambiguousbecause it is unclear what kind of social norms, if any, are more likely to de-velop. The presence of coworkers may decrease employee theft if cowork-ers monitor each other and promote integrity in the workplace, but mayincrease theft if coworkers collude against the firm and/or if coworkerpresence makes the identification of thieves more challenging, thus pro-viding greater opportunity for dishonest employees to steal (Hollinger andClark [1983]). The type of social norms that will develop and the degreeof conformity to social norms by workers will be influenced by the situa-tion, which depends to a large extent on the control environment (Zhang[2008], Tayler and Bloomfield [2011]).

Employee compensation is likely to shape the kind of social norms de-veloped in the workplace by influencing the work group’s perceived fair-ness of treatment by management (Akerlof [1982], Fischer and Huddart[2008]). According to Feldman [1984, p. 47], “ . . . if the work group feelsthat management is supportive, group norms will develop that facilitate—infact, enhance—group productivity. In contrast, if the work group feels thatmanagement is antagonistic, group norms that inhibit and impair groupperformance are much more likely to develop.” Consistent with the abovearguments, gift exchange and reciprocity models suggest that firms pay-ing relatively high wages will induce worker loyalty, which, in turn, willlead employees to interact with their peers in ways that increase the qual-ity and productivity of their output (Akerlof [1984], Fehr and Gachter

8 A social norm is defined as: “(1) a behavioral regularity; that is (2) based on a sociallyshared belief of how one ought to behave; which triggers (3) the enforcement of the pre-scribed behavior by informal social sanctions” (Fehr and Gachter [2000, p. 166]).

9 Based on the 2008 Association of Certified Fraud Examiners (ACFE) report, tips providedby coworkers, customers, or other individuals represent by far the most frequent method ofdetecting occupational fraud: 46.2% of all occupational fraud cases were detected by tips,compared to 19.4% by internal audits and 23.3% by internal controls.


[2000]). For example, in a study of machine shop workers, Burawoy[1979] finds that workers who believe they are overpaid play competitivegames with their coworkers leading to higher productivity, whereas work-ers who believe they are underpaid play destructive games leading to lowerproductivity.

In addition, the whistle-blowing literature suggests that an employee ismore willing to blow the whistle on a coworker’s deviant behavior if theemployee perceives the employer to be fair (Miceli, Near, and Schwenk[1991]). Consistent with this argument, an experimental study conductedby Zhang [2008] shows that, under a peer reporting system, the percent-age of whistle-blowing on lying peers is higher when the agents perceivethe principal as fair versus unfair. Similarly, using a field survey of 360 em-ployees in 18 fast food restaurants, Victor, Trevino, and Shapiro [1993] findthat workers who perceive themselves to be treated fairly by management(due to various factors, including pay) are more likely to blow the whistleon their coworkers who take or give away food.

Drawing on previous literature on social norms, we conjecture that theassociation between relative wages and employee theft is more pronouncedin firms with multiple coworkers. Under higher relative wages, we predictthat workers will be more likely to show positive reciprocity to their employ-ers, leading to norms of honesty and mutual monitoring that will reducetheft. For instance, under higher wages, workers should be more likely tomonitor each other and to sanction employees who steal by avoiding inter-actions with them or condemning their acts (Hollinger and Clark [1983]).Conversely, we predict that relatively lower wages will induce negative reci-procity on the part of the employees, leading to norms of dishonesty andcollusion that will increase theft. Under relatively lower wages, employeeswho observe their coworkers steal should be more likely to rationalize andcover up their coworkers’ misconduct, or to steal from the company justlike their coworkers.

In addition to the above argument based on the social norms literature,the “shirking model” also suggests an interaction effect between coworkerpresence and relative wages on employee theft. According to this theory,under higher relative wages, coworkers do not want to lose their jobs, andtherefore are more likely to report coworker theft to management to avoidbeing blamed (or fired) for theft committed by someone else. Conversely,under relatively lower wages, workers care less about losing their jobs, sothey are more permissive about coworkers’ theft and less likely to moni-tor their coworkers in order to avoid being blamed for theft committed bysomeone else.

Based upon the discussion above, we conjecture that high relative wageswill promote a more ethical environment among coworkers and reduce em-ployee theft. Conversely, low relative wages will discourage peer monitoringand encourage collusion against the company, leading to higher levels ofemployee theft. More formally:


H2: Relative wages and coworker presence interact to affect employeetheft, such that coworker presence reduces (increases) employeetheft when relative wages are higher (lower).

3. Data and Research Design

3.1 DATA

To test our hypotheses, we obtain both proprietary and publicly avail-able U.S. data from five main sources: (1) 1999 data for a medium-sizedconvenience store chain (CS Chain), (2) 2004 and 2005 store-level surveydata (reporting annual data from 2003 and 2004, respectively) from the Na-tional Association of Convenience Stores (NACS), (3) property crimes datafrom the Federal Bureau of Investigation (FBI) at http://www.fbi.gov, (4)wages and unemployment data from the Bureau of Labor Statistics (BLS)at http://www.bls.gov, and (5) data relating zip codes to MSAs from Zip-CodeDownload at http://www.zipcodedownload.com.

The convenience store industry is particularly suitable for this study forseveral reasons. First, it suffers from a severe employee theft problem thatcan be adequately captured with inventory shrinkage and cash shortage.Since the convenience store industry is a low-margin, high-volume busi-ness, employee theft can cause significant damage to business performance(Bernardin and Cooke [1993]). Second, due to the limited use of incentivecompensation for hourly employees, we are able to conduct a clean test ofthe effects of relative wages. Third, the payoffs from stealing are fairly small,which makes the use of wages to mitigate employee theft less costly than itwould be in other settings such as casinos or jewelry stores. Finally, con-venience stores provide an ideal setting to examine theories on collectivebehavior, as some convenience stores employ a single worker while othersemploy two or three workers in the same shift.

The CS Chain data set provides information on average wages, cashshortage, inventory shrinkage, employee skills, labor hours, store managerturnover, and location (among other data) for all 76 stores in a medium-sized convenience store chain for the year 1999.10 We were able to findbenchmark wages and unemployment data for the year 1999 from the BLSwebsite, as well as property crimes per capita data from the FBI website, foreach of these stores. Thus, this sample comprises 76 store-year observations.

The NACS data set provides usable information on starting wages, cashshortage, store manager turnover, and location for 329 stores from 32chains for the years 2003 and 2004 (a total of 434 store-years are avail-able).11 Members of NACS, which include retail members from all across

10 Under a confidentiality agreement, we have permission to use disguised company infor-mation to ensure the anonymity of the company.

11 According to NACS, convenience stores are defined based on the following criteria: (1)store size is less than 5,000 sq. ft.; (2) off-street parking and/or convenient pedestrian access is


the United States, were asked to report store-level survey data from arandom sample of stores within their convenience store chains. The sur-vey’s response rate was about 75%. Whenever a state was over- or under-represented, the staff at NACS adjusted the survey sample so that the NACSsample was representative of different areas across the United States. Wewere able to obtain surveys with complete data for our study from an aver-age of 7 stores per chain in 2003 and 10 stores per chain in 2004. The chainsfor which we obtained these data are relatively large, with an average (me-dian) number of stores per chain equal to 277 (53) stores in 2003 and 277(54) stores in 2004.12 Matching these data with the BLS and FBI data forthe years 2003–2004 yields a final sample of 327 store-years (correspondingto 251 stores from 31 chains) for most of our analyses.

3.2 RESEARCH DESIGN—TESTING HYPOTHESIS 1

To test Hypothesis 1, which predicts a negative association between rela-tive wages and employee theft, we estimate the following model:

Theftit = β0 + β1 Relative Wagesit + β2 Employee Skillsit

+β3 Store Manager Turnover it + β4 Property Crimes Per Capitait

+β5 Unemploymentit + β6 Year 2004 + εi t (1)

We discuss each of the variables below (variable descriptions are also inthe appendix).

For the CS Chain data set, we measure Theft using both inventory shrink-age and cash shortage scaled by store sales and multiplied by 100.13 Dueto the lack of store-level inventory shrinkage information in the NACS dataset, we measure Theft using cash shortage scaled by store sales and multi-plied by 100 in this sample.14

Our main independent variable is Relative Wages. This variable is mea-sured differently for the CS Chain and NACS data sets. The CS Chain dataset provides average hourly wages for all employees for each store in the

available; (3) the store has extended hours of operation (many are open 24 hours, seven daysa week); (4) the store offers at least 500 stock-keeping units (SKUs); and (5) the store’s prod-uct mix includes grocery-type items and items from the following groups: beverages, snacks(including confectionery), and tobacco.

12 This figure is based on 23 of the 32 retail chains. Information on the number of storesper chain (i.e., chain size) was missing for 9 of the chains.

13 Inventory shrinkage occurs when the physical inventory count is lower than the booked(invoiced) inventory (which is equal to beginning inventory plus purchases minus sales andadjustments); cash shortage occurs when the physical cash count at the end of the day is lowerthan the cash register tape (Greenberg [1990], NRSS [1992]).

14 Note that care should be taken in interpreting the results, as our inventory shrinkagemeasure may capture both employee theft and customer shoplifting, and our cash shortagemeasure does not capture the skimming form of cash theft (where cash is stolen before beingrecorded in the organization) and may capture both employee theft and employees’ mistakesin recording transactions.


chain.15 Since the BLS provides aggregate data on hourly wages for cashiersin all sales organizations for each MSA, we are able to use the BLS data asa benchmark in calculating relative wages for the CS Chain data set. Thus,for this data set, Relative Wages is calculated as the difference between theaverage hourly wage for a store’s employees and the median hourly wagefor cashiers in sales organizations in the same MSA in which the store islocated (this is a similar measure to that used by Cappelli and Chauvin[1991]).16 The NACS data set, in contrast, provides starting hourly wagesfor entry-level employees for each store included in the data set.17 Sincethe data in the BLS correspond to all employees rather than entry-level em-ployees only, we had to adjust the BLS median hourly wage by multiplyingit by 0.88 to make it comparable to starting wages.18 Thus, for the NACSdata set, Relative Wages is calculated as the difference between the startinghourly wage in a store and 88% of the median hourly wage for cashiers insales organizations in the same MSA in which the store is located.19

We include several controls in the model. First, we control for EmployeeSkills in the CS Chain sample to address a potential concern that relativewages might capture employee characteristics that could be related to anemployee’s likelihood to steal. If employee skills were positively correlatedwith honesty, we would expect a negative association between employeeskills and employee theft. Employee Skills is the average performance ratingof hourly employees. Each employee is rated by its store manager based

15 The compensation of store managers is not included in this figure.16 CS Chain has a corporate policy in place to set compensation. In particular, the head-

quarters set wage scales for stores operating in different market areas. One of the factorsconsidered in setting these scales is competition. However, the company does not have anongoing policy targeting a specific level of average compensation relative to the competitorsof each individual store. Additionally, employees are eligible to receive a raise on a periodicbasis (they are eligible for three potential raises in the first year, and one raise a year in thefollowing years). The raises are based, to a large extent, on the store managers’ evaluations ofthe employees.

17 The compensation of store managers is not included in this figure.18 We use 88% of the BLS median number as a benchmark to make the benchmark compa-

rable to starting wages since average starting wages for all firms surveyed by NACS in a givenMSA are 91.1% of the average of all the BLS wages and 84.2% of the median of all the BLSwages in the corresponding MSA.

19 A potential reverse causality concern would arise if convenience stores rewarded em-ployees with wage increases or bonuses if inventory shrinkage and cash shortages were belowtarget at the end of a period. We believe this would not be a concern in our setting for threereasons: First, according to the National Association of Convenience Stores (NACS) Compen-sation Surveys for 2003 to 2005 (covering our sample period), cashiers’ compensation is basedexclusively on fixed hourly wages and does not include any form of bonuses. Second, inter-views with two managers at CS Chain confirmed that no measure of theft (or honesty) directlyaffected cashier wages or cashier wage increases. Third, our results hold even when we usestarting wages in our NACS sample. We interviewed managers of nine different conveniencestore chains analyzed in our study to see whether starting wages could be affected by assessedhonesty of new hires. Our interviews reveal that, even though convenience stores chains con-duct integrity or background checks, information from these checks is only used to determineemployment eligibility, but not used to determine starting wages.


on service and operations performance. The rating was elicited on a scaleof 1 to 5, with the highest skills corresponding to a 5 and the lowest skillscorresponding to a 1. CS Chain provided us with the average performanceratings across all the employees in each store.20 This variable is not avail-able for the NACS sample, yet the need to control for employee skills ismitigated to some extent by the fact that we use starting rather than aver-age wages to calculate relative wages in the NACS sample. Because startingwages are set at the same level for most starting employees working at agiven convenience store, they should be less influenced than average wagesby employee characteristics that could affect an employee’s likelihood tosteal.

Second, we control for store manager turnover, which is likely to exacer-bate monitoring difficulties and lead to higher theft (Detert et al. [2007]).Indeed, the National Retail Security Survey of 1992 reports that store man-ager turnover is associated with higher inventory shrinkage in the UnitedStates (NRSS [1992]). Store Manager Turnover is calculated as the total num-ber of store manager terminations divided by the total number of storemanagers per year multiplied by 100.21

Third, we control for property crimes per capita because both inven-tory shrinkage and cash shortage are likely to increase in areas with highcrime rates—employees coming from high-crime areas may have a greaterpropensity to steal, and stores in high-crime areas face a higher risk of bur-glary. Property Crimes Per Capita is measured as the property crimes per capitain 1999 (for the CS Chain data set) and 2004 (for the NACS data set) in theMSA where the store was located, as reported by the FBI.

Our fourth control takes the unemployment rate into account. Prior lit-erature provides mixed predictions on the relation between the unemploy-ment rate and employee theft. On the one hand, high unemployment ratesmay lead to lower theft as they translate into few outside employment op-tions for employees and thus low chances of re-employment if employeesget fired due to theft (Stiglitz [1974], Shapiro and Stiglitz [1984]). On theother hand, high unemployment rates indicate unfavorable economic con-ditions, which lead employees to be more concerned about their financialsituation and thus more likely to steal. Hollinger and Clark [1983], for

20 Our Employee Skills measure is an imperfect measure of skills since it is likely to be affectednot only by skill but also by effort (which, in turn, is likely to be affected by relative wages).However, our analysis will show that the correlation between Relative Wages and Employee Skillsis insignificant (see table 2), which alleviates the concern that Employee Skills captures effortrather than skill.

21 A potential problem with this control variable is that theft may lead to store managerturnover rather than the other way around. To alleviate this reverse causality concern, wereran our regressions after replacing the store manager turnover rate of year t with the storemanager turnover rate of year t−1. We were able to do this using a subsample of our NACSdata set (where we had 270 observations with two years of data available). All of the mainresults reported for NACS in our empirical results section (section 4) are robust to utilizing alagged variable for store manager turnover.


instance, document a positive association between employees’ financialconcerns and the rate of employee theft. Given these mixed predictions,we do not predict the direction of the relationship between the unemploy-ment rate and employee theft. Unemployment is measured using the 1999(for the CS Chain data set) and 2004 (for the NACS data set) unemploy-ment rates reported by the BLS for the MSA in which a store is located.

Finally, for the NACS data set, we also use a year dummy to controlfor year fixed effects, and we use robust standard errors clustered bychain to address error correlation problems from same-chain observations(Petersen 2009).

3.3 RESEARCH DESIGN—TESTING HYPOTHESIS 2

To test Hypothesis 2, which predicts an interactive effect between relativewages and coworker presence, we regress employee theft on relative wages,coworker presence, the interaction between the two, and control variables.Specifically, we estimate the following model:

Theftit = β0 + β1 Relative Wagesit + β2 Coworker Presenceit

+β3 Relative Wagesit × Coworker Presenceit + β4 Employee Skillsit

+β5 Store Manager Turnover it + β6 Property Crimes Per Capitait

+β7 Unemploymentit + β8 Year 2004 + εi t (2)

Based on the recommendations of Aiken and West [1991], we mean-center the continuous variables in the interaction terms before includingthem in the analysis to facilitate interpretation of the main effects. As be-fore, we use robust standard errors clustered by chain for the NACS data set.

We use the number of employees per hour to capture coworker presencein a given shift in a store. This variable is calculated as the total annuallabor hours divided by the total annual opening hours for each store. It isnot unusual for convenience stores to be staffed by only one employee, inwhich case there is no coworker influence. As the number of employees perhour increases, influence from coworker presence increases.

4. Empirical Results

4.1 DESCRIPTIVE STATISTICS

Table 1 provides descriptive statistics on our main variables. On average,cash shortage accounts for 0.33% of sales revenue in the CS Chain dataset (median = 0.26%) and 0.36% of sales revenue in the NACS data set(median = 0.17%). Inventory shrinkage accounts for 1.85% (median =1.67%) of sales revenue in the CS Chain sample.

In the CS Chain sample, the average hourly wage for employees is $8.25(median = $8.16). Relative wages range from an underpayment of $0.28 toan overpayment of $2.77, with a mean of $0.93 (median = $0.86). In theNACS sample, the average starting hourly wage for entry-level employees


TA

BL

E1

Des

crip

tive

Stat

istic

s

CS

Ch

ain

NA

CS

NM

ean

Std

Dev

Min

Med

ian

Max

NM

ean

Std

Dev

Min

Med

ian

Max

Cas

hSh

orta

ge(a

s%

ofsa

les)

760.

330.

290.

010.

262.

0632

70.

360.

480.

000.

172.

58Sh

rink

age

(as

%of

sale

s)76

1.85

0.87

0.22

1.67

4.89

––

––

––

Sale

s($

mil)

761.

120.

280.

641.

101.

8632

71.

600.

930.

291.

305.

30W

ages

($)

768.

250.

636.

858.

1610

.17

327

7.20

0.59

6.00

7.00

8.63

Rel

ativ

eW

ages

($)

760.

930.

59−0

.28

0.86

2.77

327

0.55

0.61

−1.5

70.

511.

83O

peni

ngH

ours

Per

Day

7623

.21

1.86

1824

2432

522

.77

2.50

12.2

924

24C

owor

ker

Pres

ence

761.

810.

411.

241.

673.

5831

42.

621.

131.

072.

276.

51Em

ploy

eeSk

ills

763.

360.

532.

353.

214.

78–

––

––

–St

ore

Man

ager

Turn

over

760.

220.

570.

000.

003.

3932

70.

250.

500.

000.

003.

00Pr

oper

tyC

rim

esPe

rC

apita

760.

030.

000.

020.

030.

0432

70.

040.

010.

000.

040.

06U

nem

ploy

men

t76

3.11

0.69

2.10

3.00

6.10

327

5.13

0.88

3.40

5.10

7.60

Th

ista

ble

pres

ents

desc

ript

ive

stat

isti

csfo

rth

em

ain

vari

able

s(a

nd

som

eof

thei

rco

mpo

nen

ts)

for

the

CS

Ch

ain

sam

ple

(n=

76)

and

the

NA

CS

sam

ple

(n=

327)

,res

pect

ivel

y.Se

eth

eap

pen

dix

for

defi

nit

ion

sof

the

mai

nva

riab

les.

Sale

sis

ann

uals

ales

ofth

est

ore

inm

illio

ns.

Wag

esis

aver

age

wag

efo

ral

lhou

rly

empl

oyee

sfo

rC

SC

hai

nan

dst

arti

ng

wag

efo

ren

try-

leve

lem

ploy

ees

for

NA

CS.

Ope

ning

Hou

rsPe

rD

ayis

the

num

ber

ofh

ours

ast

ore

isop

enfo

rea

chda

y.


is $7.20 (median = $7.00). Relative wages range from an underpayment of$1.57 to an overpayment of $1.83, with a mean of $0.55 (median = $0.51).22

The average convenience store in the CS Chain (NACS) sample uses 1.81(2.62) employees per hour and has a store manager turnover rate of 22%(25%). The average employee skill rating in the CS Chain sample is 3.36out of 5 (median = 3.21). The average value of property crimes per capitais 0.03 for the CS Chain sample in 1999 and 0.04 for the NACS sample in2004. The average unemployment rate is 3.11% for the CS Chain sample in1999 and 5.13% for the NACS sample in 2004. Generally speaking, the keyvariables are comparable across the two samples.

Panels A and B of table 2 present the Pearson correlations between themain variables for the CS Chain sample and the NACS sample, respectively.Consistent with our hypotheses, we find that both inventory shrinkage andcash shortage are negatively and significantly correlated with relative wagesin the CS Chain, and that cash shortage is negatively and significantly cor-related with relative wages in the NACS sample. In the CS Chain sample,we also find that cash shortage is positively correlated with unemployment,while, in the NACS sample, we find that cash shortage is negatively corre-lated with coworker presence and positively correlated with store managerturnover and property crimes per capita.

4.2 RELATIVE WAGES AND EMPLOYEE THEFT

Table 3 presents the results for our test of H1. Hypothesis 1 predicts a neg-ative association between relative wages and employee theft, so we expectthe coefficient on Relative Wages to be negative. Recall that our dependentvariable, Theft, is measured by inventory shrinkage and cash shortage in theCS Chain sample, and by cash shortage only in the NACS sample. Consis-tent with H1, the coefficient on Relative Wages is significantly negative forall three measures of employee theft (β1 = −0.452, t = −2.78 for CS ChainInventory Shrinkage; β1 = −0.116, t = −2.15 for CS Chain Cash Shortage;β1 = −0.171, t = −2.13 for NACS Cash Shortage). These results providestrong support for H1.

Most of the coefficients on the control variables are consistent with ourexpectations. We find a significantly negative coefficient on Employee Skills(β2 = −0.253, t = 1.43) when we use inventory shrinkage to measure theftin the CS Chain, consistent with our expectation that “better-quality” em-ployees are less likely to steal. We also find a significantly positive coefficienton Store Manager Turnover (β3 = 0.088, t = 1.57) when we use cash short-age to measure theft in the CS Chain, consistent with our expectation that

22 Benefits account for a small percentage of total compensation in the convenience storeindustry. According to compensation surveys conducted by the National Association of Con-venience Stores (different from the surveys used for our analysis, described in section 3),benefits accounted for about 13% of total cashiers’ compensation in 2003 and 2004 in theconvenience store industry. These data could not be incorporated into our analysis, since theywere not available for our sample firms.


TA

BL

E2

Cor

rela

tion

Mat

rice

s

Pan

elA

:CS

Ch

ain

V1

V2

V3

V4

V5

V6

V7

V8

V1:

Cas

hSh

orta

ge1.

000

V2:

Shri

nkag

e0.

493∗∗

∗1.

000

V3:

Rel

ativ

eW

ages

−0.2

52∗∗

−0.2

92∗∗

1.00

0V

4:C

owor

ker

Pres

ence

0.03

60.

169

0.27

9∗∗1.

000

V5:

Empl

oyee

Skill

s−0

.068

−0.1

38−0

.010

0.01

61.

000

V6:

Stor

eM

anag

erTu

rnov

er0.

160

0.13

10.

023

−0.0

580.

050

1.00

0V

7:Pr

oper

tyC

rim

esPe

rC

apita

0.09

7−0

.035

−0.1

86−0

.145

−0.0

32−0

.151

1.00

0V

8:U

nem

ploy

men

t0.

285∗∗

0.14

1−0

.130

−0.0

40−0

.000

−0.0

700.

621∗∗

∗1.

000

Pan

elB

:NA

CS

V1

V2

V3

V4

V5

V6

V1:

Cas

hSh

orta

ge1.

000

V2:

Rel

ativ

eW

ages

−0.2

06∗∗

∗1.

000

V3:

Cow

orke

rPr

esen

ce−0

.268

∗∗∗

0.07

21.

000

V4:

Stor

eM

anag

erTu

rnov

er0.

093∗

0.07

5−0

.152

∗∗∗

1.00

0V

5:Pr

oper

tyC

rim

esPe

rC

apita

0.16

7∗∗∗

−0.0

72−0

.313

∗∗∗

0.15

8∗∗∗

1.00

0V

6:U

nem

ploy

men

t0.

006

−0.0

860.

240∗∗

∗−0

.068

−0.0

331.

000

Th

ista

ble

pres

ents

Pear

son

corr

elat

ion

sbe

twee

nth

em

ain

vari

able

sas

defi

ned

inth

eap

pen

dix.

Pan

elA

pres

ents

Pear

son

corr

elat

ion

sfo

rth

eC

SC

hai

nda

tase

tan

dpa

nel

Bpr

esen

tsPe

arso

nco

rrel

atio

ns

for

the

NA

CS

data

set.

∗ ,∗∗

,an

d∗∗

∗in

dica

teth

atth

eco

rrel

atio

nis

sign

ifica

nta

tth

e0.

10,0

.05,

and

0.01

leve

l,re

spec

tive

ly.


T A B L E 3Relation Between Relative Wages and Employee Theft

Theftit = β0 + β1 Relative Wagesit + β2 Employee Skillsit + β3 Store Manager Turnover it

+ β4 Property Crimes Per Capitait + β5 Unemploymentit + β6 Year 2004 + εi t

CS Chain CS Chain NACS CashPredicted Shrinkage Cash Shortage Shortage

Intercept ? 3.587∗∗∗ 0.413 0.201(3.63) (1.26) (0.55)

Relative Wages – −0.452∗∗∗ −0.116∗∗ −0.171∗∗

(−2.78) (−2.15) (−2.13)Employee Skills – −0.253∗ −0.045 –

(−1.43) (−0.77)Store Manager Turnover + 0.196 0.088∗ 0.086

(1.16) (1.57) (0.96)Property Crimes Per Capita + −55.489 −11.136 5.082

(−1.69) (−1.02) (1.22)Unemployment ? 0.327∗ 0.150∗∗ −0.006

(1.86) (2.57) (−0.12)Year 2004 ? – – 0.096

(1.07)Number of Observations 76 76 327Adjusted R2 0.114 0.121 0.080

The t-statistics are reported in parentheses. ∗, ∗∗, and ∗∗∗ denote that the coefficients are significant atthe 0.1, 0.05, and 0.01 levels, respectively, based on one-tailed tests for directional predictions and two-tailedtests otherwise. See the appendix for variable definitions. The t-statistics in the NACS data set are estimatedusing chain-clustered standard errors. We include a year dummy (which is equal to “1” if the year is 2004,and “0” otherwise) for the NACS data to control for year fixed effect.

higher store manager turnover would result in less monitoring and higheremployee theft. We find a significantly positive coefficient on Unemploymentin the CS Chain (β5 = 0.327, t = 1.86 for inventory shrinkage and β5 =0.150, t = 2.57 for cash shortage), consistent with economic concerns lead-ing to higher theft. Unexpectedly, when we use CS Chain inventory shrink-age as the dependent variable, we find a negative coefficient on PropertyCrimes Per Capita (β4 = −55.5, t = −1.69). A potential explanation for thisresult is that firms use better control systems (such as monitoring cameras)in areas with higher property crimes per capita, which mitigate both cus-tomer shoplifting and employee theft.

We conduct a cost-benefit analysis to see whether the dollar savings fromreductions in employee theft justify the costs of relatively higher wages. Wefirst translate the reductions in inventory shrinkage and cash shortage at-tributable to relative wages into dollar amounts, and then compare thesebenefits with the costs of increasing wages. We use CS Chain as an exam-ple since this data set includes both inventory shrinkage and cash shortage,thus allowing us to calculate the full benefits of relative wages in terms ofreducing observable proxies for employee theft. Given the coefficients onRelative Wages in the CS Chain inventory shrinkage regression (−0.452) andthe CS Chain cash shortage regression (−0.116) (see table 3), and giventhe average annual sales of $1.12 million for a CS Chain store (see table 1),


ceteris paribus, a one-dollar increase in hourly wages (and hence relativewages) leads to annual benefits of $6,362 ( = (0.452 + 0.116) /100∗ $1.12million). Given that a CS Chain store’s average annual labor hours are15,080 and that employers should pay Social Security, Medicare, federal un-employment taxes, and state taxes accounting for about 8% of labor costs,a one-dollar increase in hourly wages would cost approximately $16,285(15,080 × 1.08). This analysis suggests that, although the benefit of reduc-ing employee theft accounted for by cash shortage and inventory shrinkagedoes not completely cover the cost of paying a wage premium, it does ac-count for 39% of the cost of a wage increase. In addition, our proxies foremployee theft may not capture all instances of theft, and our analysis doesnot capture other significant benefits that higher employee wages may con-vey, such as higher employee effort or reductions in turnover costs.23 There-fore, our analysis is likely to underestimate the full benefit from higherwages. If the other benefits from wage increases translate into at least 61%of the cost of the wage increase, an employer may find it economically ben-eficial to raise employee wages.

Next, we explore whether the relationship between relative wages andemployee theft is driven by the overpayment subsample, the underpaymentsubsample, or both. While the “reciprocity” model described by Fehr andGachter [2000] suggests a role of overpayment in mitigating employee mis-conduct due to employees’ positive reciprocity to their employers, some re-searchers suggest that overpaid employees are more likely to adjust theirself-perceptions of what they are worth upwards than to reciprocate totheir firms (Akerlof and Yellen [1990], Gneezy and List [2006]). Thus, astronger effect of underpayment relative to overpayment on employee theftwould be consistent with the argument that the link between relative wagesand employee theft is primarily driven by employees’ retaliation againsttheir employers for inequitable compensation (as suggested by Greenberg’s[1990] study). Conversely, a stronger effect of overpayment relative to un-derpayment on employee theft would suggest that the link between relativewages and employee theft is mainly driven by employees’ desire to recip-rocate to a fair and generous employer and/or to retain high-paying jobswhere they feel valued.

We run the regression specified in equation (1) separately for the over-payment and underpayment subsamples.24 We are only able to do this anal-ysis for the NACS data set because there is only one observation in theunderpayment subsample for the CS Chain data set. The results from theseregressions are summarized in table 4. We find a significantly negative coef-ficient on Relative Wages (Coefficient = −0.306, t = −2.07) in the subsampleof firms where relative wages are positive, and a negative but insignificant

23 For instance, our measure of relative pay is negatively related to employee turnover inboth of our samples, though the correlation is significant only in the CS Chain sample (ρ =−0.31, p-value = 0.008).

24 Our sample did not include any cases where relative wages were equal to zero.


T A B L E 4Relation Between Relative Wages and Employee Theft for the Positive Relative Wages and Negative

Relative Wages SubsamplesTheftit = β0 + β1 Relative Wagesit + β2 Store Manager Turnover it + β3 Property Crimes Per Capitait

+ β4 Unemploymentit + β5 Year 2004 + εi t

NACS Cash Shortage

Positive Relative Negative RelativePredicted Wages Subsample Wages Subsample

Intercept ? 0.198 0.200(0.49) (0.58)

Relative Wages – −0.306∗∗ −0.223(−2.07) (−1.26)

Store Manager Turnover + 0.109 −0.116(1.06) (−0.79)

Property Crimes Per Capita + 4.625 6.888(1.01) (1.25)

Unemployment ? 0.026 −0.048(0.38) (−1.12)

Year 2004 ? 0.070 0.167∗

(0.68) (1.89)Number of Observations 269 58Adjusted R2 0.118 0.134

The t-statistics in parentheses are estimated using robust standard errors, clustered by chain. ∗ and ∗∗denote that the coefficients are significant at the 0.1 and 0.05 levels, respectively, based on one-tailed testsfor directional predictions and two-tailed tests otherwise. See the appendix for definitions of the mainvariables. We include a year dummy (which is equal to “1” if the year is 2004, and “0” otherwise) to controlfor year fixed effects.

coefficient on Relative Wages (Coefficient = −0.223, t = −1.26) in the sub-sample of firms where relative wages are negative. These results suggest thatthe link between relative wages and employee theft is predominantly drivenby overpayment. Our CS Chain results are also robust to dropping the onlycase of underpayment in that sample. These results suggest that overpaidemployees do reciprocate to their employers for generous compensation,which provides evidence against the argument in prior literature that over-paid workers often rationalize away their overpayments. A possible expla-nation for the lack of results for underpayment firms is that the tests forthis subsample of firms may lack power, not only because the sample sizeis limited, but also because employees’ wages are close to minimum wages,and thus the magnitude of underpayment may be truncated. Another pos-sibility is that there is a ceiling to how much employees are willing to stealfrom their employers, that is, employees may be willing to engage in pettytheft to retaliate against their employers but they may be reluctant to steala large amount of money.

4.3 INTERACTION EFFECTS BETWEEN RELATIVE WAGES AND COWORKERPRESENCE

Hypothesis 2 posits an interaction effect between relative wages andcoworker presence on employee theft. To test this hypothesis, we estimateequation (2) for our three measures of employee theft. The results are


T A B L E 5Effects of Relative Wages, Coworker Presence, and the Interaction Between Relative Wages and Coworker

Presence on Employee TheftTheftit = β0 + β1 Relative Wagesit + β2 Coworker Presenceit + β3 Relative Wagesit × Coworker Presenceit

+ β4 Employee Skillsit + β5 Store Manager Turnover it + β6 Property Crimes Per Capitait

+ β7 Unemploymentit + β8 Year 2004 + εi t

CS Chain CS Chain NACS CashPredicted Shrinkage Cash Shortage Shortage

Intercept ? 3.231∗∗∗ 0.301 0.128(3.52) (0.94) (0.33)

Relative Wages – −0.455∗∗∗ −0.122∗∗ −0.143∗

(−2.69) (−2.07) (−1.69)Coworker Presence ? 0.724∗∗∗ 0.097 −0.098∗∗∗

(2.94) (1.13) (−4.06)Relative Wages × Coworker Presence – −0.453∗∗ −0.039 0.040

(−1.83) (−0.46) (0.91)Employee Skills – −0.300∗∗ −0.050 –

(−1.76) (−0.83)Store Manager Turnover + 0.178 0.088∗ 0.075

(1.09) (1.54) (0.85)Property Crimes Per Capita + −51.46 −10.172 2.519

(−1.62) (−0.92) (0.58)Unemployment ? 0.332∗ 0.148∗∗ 0.011

(1.97) (2.51) (0.21)Year 2004 ? – – 0.095

(1.01)Number of Observations 76 76 314Adjusted R2 0.198 0.112 0.120

The t-statistics are reported in parentheses. ∗, ∗∗, and ∗∗∗ denote that the coefficients are significant atthe 0.1, 0.05, and 0.01 levels, respectively, based on one-tailed tests for directional predictions and two-tailedtests otherwise. See the appendix for variable definitions. Variables used in the interaction terms are mean-centered before being included in the analysis. The t-statistics in the NACS sample are estimated usingchain-clustered standard errors. We include a year dummy (which is equal to “1” if the year is 2004, and “0”otherwise) for the NACS data set to control for the year fixed effect.

summarized in table 5. A negative coefficient on the interaction term be-tween Relative Wages and Coworker Presence (β3) would be consistent withH2, which predicts that the effect of coworker presence on employee theftis contingent on relative wages.

As shown in table 5, consistent with H2, the coefficient on the interac-tion term between Relative Wages and Coworker Presence (β3) is significantlynegative (β3 = −0.453, t = −1.83) when we use inventory shrinkage as ourmeasure of employee theft. We also find a significantly negative main effectof relative wages (β1 = −0.455, t = −2.69). Interestingly, we find a signif-icantly positive main effect of coworker presence (β2 = 0.724, t = 2.94),suggesting that inventory shrinkage is higher when more coworkers arepresent.25 The inventory shrinkage results (illustrated in figure 1) suggestthat relatively higher wages mitigate potential collusion among coworkers.

25 Our result that inventory shrinkage is higher when more coworkers are present couldbe due to collusion among coworkers or due to the fact that the presence of more employees


FIG

1.—

Inte

ract

ion

effe

ctof

rela

tive

wag

esan

dco

wor

ker

pres

ence

onin

ven

tory

shri

nka

ge.


When we use cash shortage as our measure of employee theft, the maineffect of relative wages on employee theft continues to be significantly neg-ative. However, we do not find significant coefficients on the interactionterms between relative wages and coworker presence (β3 = −0.039, t =−0.46 for CS Chain; β3 = 0.040, t = 0.91 for NACS). Instead, we find asignificantly negative main effect of coworker presence in the NACS sam-ple (β2 = −0.098, t = −4.06). We conjecture that a possible reason for thisresult is that coworkers consider cash theft to be a more serious crime thaninventory theft, and therefore exert stronger social incentives against cashtheft than inventory shrinkage. This is consistent with findings from themarketing and psychology literature that people behave less honestly withrespect to other mediums compared with money (e.g., Mazar, Amir, andAriely [2008]) because mediums other than money (e.g., inventory) allowpeople to reinterpret their theft in a more self-serving manner in order tomaintain their self-concept.26 Since inventory shrinkage is subject to lesssocial sanction from coworkers than cash shortage, inventory shrinkage islikely to be more heavily influenced by group norms that may emerge fromrelative pay than cash shortage.

Taken together, we find partial support for H2. Specifically, we findthat relative wages moderate the effect of coworker presence on inventoryshrinkage, but do not moderate the effect of coworker presence on cashshortage.

4.4 ROBUSTNESS CHECKS

We conducted additional analyses to assess the sensitivity of our results.First, in addition to the variables included in the main regressions, we

add control variables that capture the socioeconomic and monitoring en-vironment in the store as well as employee characteristics. These variablesare not included in the main regressions because they are available for onlyone data set or for only a fraction of the observations in both data sets (theappendix provides detailed descriptions of the variables considered). Theresults are presented in table 6.

increases the difficulty of detecting theft, providing more opportunity for dishonest employeesto steal. To disentangle these two alternative explanations, we argue that collusion should be-come more difficult and thus weaker as the number of coworkers increases, but the difficultyof detecting coworker theft should increase as the number of coworkers sharing a shift in-creases. Thus, if the result is primarily driven by collusion among coworkers, we would expectto see a reduction in the main effect of coworker presence on inventory shrinkage as the num-ber of coworkers increases. To test this prediction, we add a quadratic term of Coworker Presenceinto our equation (excluding the interaction term between Relative Wages and Coworker Pres-ence) and find that inventory shrinkage is positively and significantly associated with CoworkerPresence but negatively and significantly associated with the squared value of Coworker Presence.This result is consistent with our result being primarily driven by collusion.

26 Consistent with this theory, Dubner and Levitt’s [2004, p. 62] tale of the economist-turned-bagel man documents an interesting statistic: “the same people who routinely stealmore than 10% of his bagels almost never stoop to stealing his money box—a tribute to thenuanced social calculus of theft.”


T A B L E 6Replication of Results in Table 3 (Test of H1) and Table 5 (Test of H2) After Including Additional

Control Variables

Panel A: CS ChainTest of H1 Test of H2

CS Chain CS Chain CS Chain CS ChainPredicted Shrinkage Cash Shortage Shrinkage Cash Shortage

Intercept ? 2.871∗∗∗ −0.157 2.826∗∗∗ −0.203(2.63) (−0.47) (2.75) (−0.61)

Relative Wages – −0.280∗ −0.063 −0.315∗∗ −0.064(−1.60) (−1.17) (−1.70) (−1.06)

Coworker Presence ? 0.623∗∗∗ 0.035(2.58) (0.44)

Relative Wages ×Coworker Presence

– −0.396∗∗ −0.026

(−1.66) (−0.33)Employee Skills – −0.226 −0.018 −0.282∗ −0.022

(−1.30) (−0.34) (−1.68) (−0.39)Employee Experience – 0.001 0.002 0.003 0.002

(0.18) (0.89) (0.39) (0.90)Store Manager

Turnover+ 0.261∗ 0.084∗ 0.253∗ 0.083∗

(1.51) (1.58) (1.47) (1.49)Employee Turnover + 0.208∗∗∗ 0.100∗∗∗ 0.183∗∗∗ 0.098∗∗∗

(3.11) (4.82) (2.81) (4.64)Property Crimes Per

Capita+ −54.169 −7.127 −53.246 −7.128

(−1.69) (−0.72) (−1.71) (−0.71)Unemployment ? 0.329∗ 0.152∗∗∗ 0.343∗∗ 0.153∗∗∗

(1.94) (2.91) (2.09) (2.88)Income Per Capita −0.005 0.001 −0.005 0.001

(−0.75) (0.63) (−0.88) (0.60)Number of

Observations76 76 76 76

Adjusted R2 0.210 0.323 0.269 0.305

Panel B: NACSTest of H1 Test of H2NACS Cash NACS Cash

Predicted Shortage Shortage

Intercept ? −0.389 −0.239(−1.13) (−0.66)

Relative Wages – −0.109∗∗ −0.048(−2.02) (−0.75)

Coworker Presence ? −0.062∗∗∗

(−2.69)Relative Wages × Coworker Presence – 0.006

(0.23)Employee Experience – −0.161 −0.127

(−0.85) (−0.70)Employee Age ? 0.989∗∗ 0.601∗

(2.28) (1.34)

(Continued)


T A B L E 6 —Continued

Panel B: NACSTest of H1 Test of H2

NACS Cash NACS CashPredicted Shortage Shortage

Store ManagerTurnover

+ 0.066 0.049

(1.04) (0.87)Employee Turnover + 0.092∗∗∗ 0.094∗∗∗

(3.81) (5.20)Corporate Monitoring

Spending– −8.713 −9.935∗∗

(−1.26) (−1.49)Property Crimes Per

Capita+ −0.357 −0.274

(−0.22) (−0.20)Unemployment ? 0.005 0.009

(0.22) (0.39)Income Per Capita – −0.002 −0.0002

(−0.77) (−0.08)Year 2004 ? 0.035 0.028

(0.50) (0.42)Number of

Observations133 128

Adjusted R2 0.256 0.313

The t-statistics are reported in parentheses. ∗, ∗∗, and ∗∗∗ denote that the coefficients are significant atthe 0.1, 0.05, and 0.01 levels, respectively, based on one-tailed tests for directional predictions and two-tailedtests otherwise. See the appendix for variable definitions. Variables used in the interaction terms are mean-centered before being included in the analysis. The t-statistics in the NACS sample are estimated usingchain-clustered standard errors. We include a year dummy (which is equal to “1” if the year is 2004, and “0”otherwise) for the NACS data set to control for the year fixed effect.

In both the CS Chain dataset and the NACS data set, we add EmployeeTurnover because: (1) higher employee turnover results in difficulty ofmonitoring and thus greater opportunities to misbehave, which may in-crease employee theft (Hollinger and Clark [1983]), and (2) employeeswho expect to leave their jobs may be more likely to steal from the firm(Hollinger and Clark [1983], Thoms et al. [2001]). Notice, however, em-ployee turnover should be interpreted with caution for two reasons: (1) likeemployee theft, turnover is an outcome of relative wages (Stiglitz [1974],Campbell III [1993]); (2) employees that are caught stealing will likely beasked to leave the firm, thus turnover might be endogenously related totheft.27

27 We did not include employee turnover in the main analysis because previous literaturehas shown both that relative wages affect employee turnover (Levine [1993]) and that em-ployee turnover can be an antecedent of employee theft (Hollinger and Clark [1983], Thomset al. [2001]). As our analysis captures the total effect of relative wages on employee theft,it likely has already incorporated the indirect effect of relative wages on employee theft viaemployee turnover. Further analyses confirm this mediating effect of employee turnover onthe relationship between relative wages and theft: (a) Untabulated results show that relativewages have a negative effect on employee turnover, after controlling for other control variables


We also control for Employee Experience for two purposes: First, we use em-ployee experience to control for the possibility that some inventory shrink-age and cash shortage were caused by mistakes due to inexperience. Sec-ond, we control for employee experience since our interviews revealed thatthis variable could be a determinant of wages in both samples. Interviewswith two CS Chain managers reveal that, while starting wages at this chainare the same for all employees at any given store, wage increases are deter-mined based on the employee’s experience and performance (captured bythe “employee skills” measure). A store manager explained: “Their wagesget higher the longer they have worked with us . . . We have performanceevaluations for the cashiers every three months. If they get a bad perfor-mance evaluation, we will hold the raise, but even if they get good perfor-mance, it doesn’t mean that they will get a raise because they will only geta raise every year.” Interviews with store managers of convenience stores inour NACS sample reveal that most starting employees from a given storereceive the same wage when they are first hired, but managers from someconvenience stores indicated that they occasionally make small adjustmentsto the starting wage based on the employee’s experience. To the extent thatemployee experience is positively associated with honesty, we expect expe-rience to be negatively related with theft.

For both the NACS and the CS chain samples, we also control for IncomePer Capita to account for economic pressures that may affect the behavior ofemployees. In the NACS Chain data set, we also control for Corporate Mon-itoring Spending and Employee Age. Monitoring spending should deter em-ployee theft by increasing the probability of theft detection and in turn theexpected costs of stealing (Nagin et al. [2002]). Regarding employee age,Hollinger and Clark [1983] suggest that younger employees are more likelyto steal. However, a recent study suggests that, while younger employees aremore likely to engage in merchandise theft, their older peers are morelikely to steal cash based on financial needs (Fischer and Green [1998,p. 298]).

We obtain our employee turnover measure from internal records at CSChain and the store-level survey from NACS, respectively. Our employeeexperience measure at CS Chain is obtained from its internal records onemployee tenure. The income per capita measure comes from the 2000U.S. Census. The corporate monitoring spending, employee experience,

(store manager turnover, unemployment, and a proxy for store competition defined for theCS Chain sample as the number of competitors within one mile divided by population; andfor the NACS sample, as an indicator variable that is equal to “1” if a store is in an urban loca-tion, and “0” otherwise), yet this relationship is significant only in the CS Chain sample (Coef.= −0.78, t-stat. = −2.84); and (b) Table 6 shows that employee turnover has a significantlypositive effect on theft after controlling for relative wages and other control variables. Theslightly weaker results of relative wages on theft in table 6 after we include employee turnoverare consistent with employee turnover partially mediating the relationship between relativewages and employee theft.


and employee age measures used in the NACS sample come from a chain-level survey (rather than the store-level survey described in section 3) fromNACS. The addition of these controls results in a CS Chain sample of 76stores, and a NACS sample of 133 store-years (128 store-years when we addcoworker presence to the analyses).

As shown in table 6, our main results for both the CS Chain sample andthe NACS sample are robust to the inclusion of all of these variables, yetthe effect of relative wages on cash shortage becomes insignificant in theCS Chain sample. Additionally, we find that there are significantly positiveassociations between employee turnover and both inventory shrinkage andcash shortage in both samples. We find that employee experience and in-come per capita are insignificantly related to both inventory shrinkage andcash shortage in both samples. In the NACS sample, we find a positive as-sociation between employee age and cash shortage, suggesting that olderemployees are more likely to steal cash than younger employees. We alsofind that corporate monitoring spending is significantly negatively relatedto cash shortage, consistent with our expectation that stronger corporatemonitoring reduces employee theft. 28

Second, we estimate all the regression models (tables 3 to 5) using alter-native wage benchmarks. We consider two sets of alternative measures:

(a) We measure relative wages as the difference between the averagehourly wage for a store’s employees and the mean (instead of median)of the benchmark wages (hourly wages for sales cashiers in the sameMSA for CS Chain and 88% of hourly wages for sales cashiers in thesame MSA for NACS). These alternative specifications do not substan-tively change any of our results in tables 3, 4, or 5.

(b) We consider an alternative benchmark for the NACS Chain sample,using the median starting hourly wages of all the convenience storesin the NACS data set that were in the same region as the store29 ratherthan 88% of the median hourly wages of sales cashiers reported inthe BLS website. Using this alternative benchmark does not changethe results presented in any of the tables, except that the negativemain effect of relative wages on cash shortage becomes insignificantif we introduce the interaction between relative wages and coworkerpresence in the regression.

Third, we consider alternative definitions for our employee theft proxies.Instead of scaling cash shortages and inventory shrinkage by sales, we scalethem by store size (measured as store square footage). The advantage of

28 We estimate the variance inflation factor for each variable in each of the six regressionspresented in table 6. The variance inflation factors associated with the variables reported ineach of the regressions range from 1.07 to 1.83 in the CS Chain sample and from 1.13 to 2.43in the NACS sample. These results mitigate multicollinearity concerns.

29 The NACS compensation survey defines six regions based on geographical location inthe United States (i.e., Northeast, Southeast, Midwest, Southwest, Plains, and West).


using store size rather than sales to scale our theft measures is that storesize cannot be affected by employee behavior while sales might be affectedby employees based on whether they receive higher or lower pay. Usingthis alternative definition of our dependent variable yields similar results tothose presented in tables 3 to 5.

Fourth, for both the CS Chain and the NACS data sets, we eliminate in-fluential observations, that is, observations for which the Cook’s Distance ishigher than 4/N (where N is the number of observations in the sample).We obtain similar results in all of our regressions (equivalent to tables 3to 5), except that the negative association between relative wages and em-ployee theft is significant not only for the overpayment but also for theunderpayment subsample in table 4.

Finally, for the NACS data set, we use robust standard errors double-clustered by store and by chain to address correlation problems from same-store observations across years as well as same-chain observations (Petersen[2009]). The results are qualitatively similar to those reported in tables 3to 5.

5. Conclusion

The objective of this study is to examine the extent to which relativewages mitigate employee theft. Using two complementary store-level datasets from the convenience store industry, we find that relative wages areassociated with lower employee theft. We also find that employee theft de-creases in the magnitude of overpayment but does not increase in the mag-nitude of underpayment. Finally, we find that relative wages and coworkerpresence interact to influence theft (as measured by inventory shrinkage)such that coworker presence is associated with lower inventory shrinkagewhen relative wages are higher. In summary, results of our study shed lighton the impact of compensation practices on shaping employee honesty andethical norms in organizations.

Our research contributes to the efficiency wages and management con-trol literature by identifying an additional benefit of efficiency wages thathas not been fully explored in prior literature. While previous studies havefocused on the effect of relatively higher wages on employee effort and/orturnover, we document the effect of relatively higher wages on employeetheft. In doing so, we provide an additional argument for why some em-ployers offer wage premiums and suggest an alternative honesty-inducingmechanism to other internal control mechanisms studied in the account-ing literature. Our study also contributes to the literature on social normsby showing that the presence of coworkers is more likely to promote eth-ical norms of behavior when employees are paid relatively higher wages.Consistent with insights from recent experimental studies conducted by ac-counting researchers (e.g., Tayler and Bloomfield [2011], Hannan, Towry,and Zhang [2011]), we show that compensation practices, as part of theformal control systems of a company, can shape the social context of thework group, which, in turn, influences employee theft.


The results of our study have important practical implications for man-agers. Understanding the full benefits of wage premiums should help man-agers determine employee wage levels. However, previous research pro-vides little guidance to managers regarding the benefit of wage premiumsin curbing employee theft. Our research provides systematic empirical evi-dence that wage premiums do play a role in reducing employee theft andfostering more ethical norms within an organization. These results lendsupport to Pfeffer’s [1994] argument that paying higher-than-market wageshelps companies effectively manage their employees as a way to producesustainable competitive advantage. Takeaways from our study are likely toapply to other types of retailers (such as restaurants, department stores,drug stores, etc.) and to service or consumer products firms with similarmonitoring environments, where the payoffs from stealing are not dispro-portionately high relative to potential wage premiums (e.g., casinos).

An interesting result of our study is that the benefit of reducing theamount of employee theft accounted for by cash shortage and inventoryshrinkage does not, by itself, outweigh the cost of paying a wage premium.Yet, it accounts for about 39% of the cost of a wage increase. An employermay find it economically beneficial to raise employee wages if other bene-fits from wage increases (e.g., reduced employee turnover, greater effort)translate into at least 61% of the cost of the wage increases. Our findingalso suggests that the benefits from increasing employee pay are likely tobe greater if firms use multiple employees per shift to staff their stores be-cause higher wages also induce more ethical norms among the coworkers.

The results of this study should be interpreted with some caveats in mind.As we mention above, neither inventory shrinkage nor cash shortage mea-sure employee theft without error. Moreover, both measures capture onlythe suspected level of theft, not the actual level of employee theft. Never-theless, these measures provide reasonable proxies for employee theft inthe convenience store industry and our study advances our understandingof the impact of employee pay on workplace honesty. Additionally, due todata restrictions, we obtained our relative wages measures using the pre-vailing wages in the store’s MSA as a benchmark. If finer-grained data (e.g.,prevailing wage at the zip code level) became available, future researchcould examine more precise measures of relative wages. Also, employeesmay compare their wages to those of comparable employees within thesame organization. As different psychological mechanisms may be activatedwith different reference points, future research can examine the effects ofrelative wages using other benchmarks. Finally, although we focus on rela-tive wages as one approach to reduce employee theft in this study, there arealternative ways to curb employee theft. For example, in a setting where em-ployees are provided with incentive pay schemes, tying employee bonusesto profits or the levels of inventory shrinkage or cash shortage may reduceemployee theft. Future research can explore other compensation mecha-nisms companies can use to control employee theft.


APPENDIXVariable Definitions

Measures CS Chain NACS

Dependent VariablesTheft

� Shrinkage� Cash Shortage

(1) Inventory shrinkage scaledby store sales and multipliedby 100

Cash shortage scaled by storesales and multiplied by 100

(2) Cash shortage scaled bystore sales and multiplied by100

Main Independent VariablesRelative Wages Main analysis: Main analysis:

Difference between theaverage hourly wage foremployees in a store and themedian hourly wage forcashiers in salesorganizations in the sameMSA in which the store islocated

Difference between thestarting hourly wage foremployees in a store and88% of the median hourlywage for cashiers in salesorganizations in the sameMSA in which the store islocated (88% is theadjustment required tomake median hourly wagesin the MSA comparable tostarting wages)

Robustness tests: Robustness tests:Difference between the

average hourly wage foremployees in a store and themean hourly wage forcashiers in salesorganizations in the sameMSA in which the store islocated

� Difference between thestarting hourly wage foremployees in a store and88% of the mean hourlywage for cashiers in salesorganizations in the sameMSA in which the store islocated

� Difference between thestarting hourly wage in astore and the median startinghourly wage of all theconvenience stores in thesame region that the store islocated (NACS defines sixgeographical regions:Northeast, Southeast,Midwest, Southwest, Plains,and West)

Coworker Presence Total annual labor hoursdivided by the total annualopening hours for eachstore

Total annual labor hoursdivided by the total annualopening hours for eachstore

(Continued)


APPENDIX —Continued

Measures CS Chain NACS

Controls for Employee CharacteristicsEmployee Skills Average rating of hourly

employees in the store basedon service and operations,with higher skillscorresponding to higherratings on a 1 to 5 scale

Not available

Employee Experience(Robustness test)

Average number of months ofexperience (tenure) of thehourly employees in a store

Percentage of employees inthe chain who have over oneyear of experience

Employee Age(Robustness test)

Not available Percentage of employees inthe chain who are over 24years old

Controls for Monitoring Difficulty and Monitoring SpendingStore Manager

TurnoverTotal number of store

manager terminationsdivided by the total numberof store managers

Total number of storemanager terminationsdivided by the total numberof store managers

Employee Turnover(Robustness test)

Total number of hourlyemployee terminationsdivided by the total numberof hourly employees at yearend for a given store

Total number of hourlyemployee terminationsdivided by the total numberof hourly employees at yearend for a given store

CorporateMonitoringSpending(Robustness test)

N/A—This is only one chain.Thus, corporate monitoringspending is implicitlycontrolled for.

Corporate security spendingdivided by the total directoperating expenses acrossall the stores of the chain.Security spending typicallyincludes monitors, alarms,security personnel, andarmored-car pick-ups.

Controls for Socioeconomic EnvironmentProperty Crimes Per

CapitaProperty crimes per capita that

occurred in 1999 in thesame MSA where the store islocated, as reported by theFBI

Property crimes per capita thatoccurred in 2004 in thesame MSA where the store islocated, as reported by theFBI

Unemployment 1999 unemployment ratesreported by the U.S. Bureauof Labor Statistics for theMSA where the store islocated

2004 unemployment ratesreported by the U.S. Bureauof Labor Statistics for theMSA where the store islocated

Income Per Capita(Robustness test)

Income per capita in 2000 inthe zip code where the storeis located

Income per capita in 2000 inthe zip code where the storeis located


REFERENCES

ADAMS, J. S. “Toward an Understanding of Inequity.” Journal of Abnormal and Social Psychology67 (1963): 422–36.

AIKEN, L. S., AND S. G. WEST. Multiple Regression: Testing and Interpreting Interactions. NewburyPark, CA: Sage Publications, 1991.

AKERLOF, G. A. “Labor Contracts as Partial Gift Exchange.” The Quarterly Journal of Economics97 (1982): 543–69.

AKERLOF, G. A. “Gift Exchange and Efficiency-Wage Theory: Four Views.” The American Eco-nomic Review 74 (1984): 79–83.

AKERLOF, G. A., AND J. YELLEN. “The Fair Wage-Effort Hypothesis and Unemployment.” Quar-terly Journal of Economics 105 (1990): 255–83.

ASSOCIATION OF CERTIFIED FRAUD EXAMINERS (ACFE). Report to the Nation. Austin, TX. Availableat: http://www.acfe.com/documents/2008-rttn.pdf, 2008.

BAKER, G. P.; K. J. MURPHY; AND M. C. JENSEN. “Compensation and Incentives: Practice vs.Theory.” Journal of Finance 43 (1988): 593–616.

BERNARDIN, H. J., AND D. K. COOKE. “Validity of an Honesty Test in Predicting Theft AmongConvenience Store Employees.” Academy of Management Journal 36 (1993): 1097–108.

BEWLEY, T. F. Why Wages Don’t Fall During a Recession. Cambridge, MA: Harvard University Press,1999.

BRICKLEY, J. A., AND F. H. DARK. “The Choice of Organizational Form: The Case of Franchis-ing.” Journal of Financial Economics 18 (1987): 401–20.

BROWN, C., AND J. MEDOFF. “The Employer Size Wage Effect.” Journal of Political Economy 97(1989): 1027–59.

BURAWOY, M. Manufacturing Consent: Changes in the Labor Process Under Monopoly Capitalism.Chicago, IL: University of Chicago Press, 1979.

CAMPBELL III, C. M. “Do Firms Pay Efficiency Wages? Evidence with Data at the Firm Level.”Journal of Labor Economics 11 (1993): 442–70.

CAMPBELL, D.; S. DATAR; AND T. SANDINO. “Organizational Design and Control Across MultipleMarkets: The Case of Franchising in the Convenience Store Industry.” The Accounting Review84 (2009): 1749–79.

CAPPELLI, P., AND W. F. CASCIO. “Why Some Jobs Command Wage Premiums: A Test of CareerTournament and Internal Labor Market Hypotheses.” The Academy of Management Journal 34(1991): 848–68.

CAPPELLI, P., AND K. CHAUVIN. “An Interplant Test of the Efficiency Wage Hypothesis.” TheQuarterly Journal of Economics 106 (1991): 769–87.

CHOW, C. W.; J. C. COOPER; AND W. S. WALLER. “Participative Budgeting: Effects of a Truth-Inducing Pay Scheme and Information Asymmetry on Slack and Performance.” The Account-ing Review 63 (1988): 111–22.

COSO (Committee of Sponsoring Organizations of The Treadway Commission). InternalControl—Integrated Framework. Vol 1, 1994.

DETERT, J. R.; L. K. TREVINO; E. R. BURRIS; AND M. ANDIAPPAN. “Managerial Modes of Influ-ence and Counterproductivity in Organizations: A longitudinal Business-Unit-Level Investi-gation.” Journal of Applied Psychology 92 (2007): 993–1005.

DUBNER, S. J., AND S. D. LEVITT. “What the Bagel Man Saw.” New York Times, June 6, 2004, p.62.

DWYER, S. “Internal C-Store Theft: Stopping the Profit Drain.” National Petroleum News 84(1992): 40–42.

EVANS III, J. H.; R. L. HANNAN; R. KRISHNAN; AND D. V. MOSER. “Honesty in Managerial Re-porting.” The Accounting Review 76 (2001): 537–59.

FEHR, E., AND A. FALK. “Wage Rigidity in a Competitive Incomplete Contract Market.” Journalof Political Economy 107 (1999): 106–34.

FEHR, E., AND S. GACHTER. “Fairness and Retaliation: The Economics of Reciprocity.” Journalof Economic Perspectives 14 (2000): 159–81.


FEHR, E.; KIRCHSTEIGER, G.; AND A. RIEDL. “Does Fairness Prevent Market Clearing? An Exper-imental Investigation.” The Quarterly Journal of Economics 108 (1993): 437–59.

FELDMAN, D. C. “The Development and Enforcement of Group Norms.” The Academy of Man-agement Review 9 (1984): 47–53.

FISCHER, P., AND S. HUDDART. “Optimal Contracting with Endogenous Social Norms.” AmericanEconomic Review 98 (2008): 1459–75.

FISCHER, R. J., AND G. GREEN. Introduction to Security, Sixth edition. Burlington, MA: ElsevierScience, 1998.

GNEEZY, U., AND J. A. LIST. “Putting Behavioral Economics to Work: Testing for Gift Exchangein Labor Markets Using Field Experiments.” Econometrica 74 (2006): 1365–84.

GREENBERG, J. “Employee Theft as a Reaction to Underpayment Inequity: The Hidden Cost ofPay Cuts.” Journal of Applied Psychology 75 (1990): 561–68.

GREENBERG, J. “The Social Side of Fairness: Interpersonal and Informational Classes of Orga-nizational Justice,” in Justice in the Workplace: Approaching Fairness in Human Resource Manage-ment, edited by R. Cropanzano Hillsdale, NJ: Erlbaum, 1993: 79–103.

GREENBERG, J. “Who Stole the Money, and When? Individual and Situational Determinantsof Employee Theft.” Organizational Behavior and Human Decision Processes 89 (2002): 985–1003.

GREENHOUSE, S. “How Costco Became the Anti-Wal-Mart.” The New York Times, July 17, 2005,p. 1.

HANNAN, R. L. “The Combined Effect of Wages and Firm Profit on Employee Effort.” TheAccounting Review 80 (2005): 167–88.

HANNAN, R. L.; J. H. KAGEL; AND D. V. MOSER. “Partial Gift Exchange in an Experimental La-bor Market: Impact of Subject Population Differences, Productivity Differences, and EffortRequests on Behavior.” Journal of Labor Economics 20 (2002): 925–51.

HANNAN, R. L.; K. TOWRY; AND Y. ZHANG. “Turning Up the Volume: An Experimental Inves-tigation of the Role of Mutual Monitoring in Tournaments.” Working paper, Georgia StateUniversity, Emory University, and Northeastern University, 2011.

HANSEN, S. C. “Designing Internal Controls: The Interaction Between Efficiency Wages andMonitoring.” Contemporary Accounting Research 14 (1997): 129–63.

HESFORD, J. W., AND J. M. PARKS. “Give and Take: Incentive Framing in Compensation Con-tracts.” Working paper, University of Akron and Washington University, 2010.

HOLLINGER, R. C., AND J. P. CLARK. Theft by Employees. Lexington, MA: Lexington, 1983.JAWORSKI, B. J., AND S. M. YOUNG. “Dysfunctional Behavior and Management Control: An

Empirical Study of Marketing Managers.” Accounting, Organizations and Society 17 (1992):17–35.

KRUEGER, A., AND L. SUMMERS. “Efficiency Wages and the Inter-Industry Wage Structure.”Econometrica 56 (1988): 259–93.

KUANG, X., AND D. MOSER. “Reciprocity and the Effectiveness of Optimal Agency Contracts.”The Accounting Review 84 (2009): 1671–94.

LEE, D., AND N. G. RUPP. “Retracting a Gift: How Does Employee Effort Respond to WageReductions?” Journal of Labor Economics 25 (2007): 725–61.

LEVINE, D. I. “What Do Wages Buy?” Administrative Science Quarterly 38 (1993): 462–83.MALCOMSON, J. M. “Unemployment and the Efficiency Wage Hypothesis.” The Economic Journal

91 (1981): 848–66.MATUSZEWSKI, L. J. “Honesty in Managerial Reporting: Is It Affected by Perceptions of Hori-

zontal Equity?” Journal of Management Accounting Research 22 (2010): 233–50.MAZAR, N.; O. AMIR; AND D. ARIELY. “The Dishonesty of Honest People: A Theory of Self-

Concept Maintenance.” Journal of Marketing Research 45 (2008): 633–44.MICELI, M. P.; J. P. NEAR; AND C. R. SCHWENK. “Who Blows the Whistle and Why?” Industrial &

Labor Relations Review 45 (1991): 113–30.MURPHY, K. R. Honesty in the Workplace. Belmont, CA: Brooks/Cole, 1993.NAGIN, D. S.; J. B. REBITZER; S. SANDERS; AND L. J. TAYLOR. “Monitoring, Motivation, and Man-

agement: The Determinants of Opportunistic Behavior in a Field Experiment.” The AmericanEconomic Review 92 (2002): 850–73.


NATIONAL RETAIL SECURITY SURVEY (NRSS). Gainesville, FL: University of Florida, 1992.NATIONAL RETAIL SECURITY SURVEY (NRSS). Gainesville, FL: University of Florida, 2005.NATIONAL RETAIL SECURITY SURVEY (NRSS). Gainesville, FL: University of Florida, 2008.NEEDLEMAN, S. E. “Businesses Say Theft by Their Workers Is Up.” The Wall Street Journal , De-

cember 11, 2008, p. B8.PETERSEN, M. A. “Estimating Standard Errors in Finance Panel Data Sets: Comparing Ap-

proaches.” Review of Financial Studies 22 (2009): 435–80.PFEFFER, J. Competitive Advantage Through People. Boston, MA: Harvard Business School Press,

1994.PFEFFER, J., AND N. LANGTON. “The Effect of Wage Dispersion on Satisfaction, Productivity,

and Working Collaboratively: Evidence from College and University Faculty.” AdministrativeScience Quarterly 38 (1993): 382–407.

ROBINSON, S. L., AND O’LEARY-KELLY, A. M. “Monkey See, Monkey Do: The Influence of WorkGroups on the Antisocial Behavior of Employees.” Academy of Management Journal 41 (1998):658–72.

SHAPIRO, C., AND J. STIGLITZ. “Equilibrium Unemployment as a Discipline Device.” AmericanEconomic Review 74 (1984): 433–44.

SHAW, H. “The Trouble with COSO.” CFO Magazine, March 15, 2006.SPEIZER, I. “An Anti-Shrinkage Strategy.” Workforce Management, 83 (2004): 52–53.STIGLITZ, J. E. “Alternative Theories of Wage Determination and Unemployment in LDC’s:

The Labor Turnover Model.” Quarterly Journal of Economics 88 (1974): 194–227.TAYLER, W. B., AND R. J. BLOOMFIELD. “Norms, Conformity, and Controls.” Journal of Accounting

Research 49 (2011): 753–90.THOMS, P.; P. WOLPER; K. S. SCOTT; AND D. JONES. “The Relationship Between Immediate

Turnover and Employee Theft in the Restaurant Industry.” Journal of Business and Psychology15 (2001): 561–77.

TOWRY, K. L. “Control in a Teamwork Environment—The Impact of Social Ties on the Effec-tiveness of Mutual Monitoring Contracts.” The Accounting Review 78 (2003): 1069–95.

VICTOR, B., L. K. TREVINO, AND D. L. SHAPIRO. “Peer Reporting of Unethical Behavior: TheInfluence of Justice Evaluations and Social Context Factors.” Journal of Business Ethics 12(1993): 253–63.

WEBB, A. “The Impact of Reputation and Variance Investigations on the Creation of BudgetSlack.” Accounting, Organizations and Society 27 (2002): 361–78.

WEISS, A. “Job Queues and Layoffs in Labor Markets with Flexible Wages.” Journal of PoliticalEconomy 88 (1980): 526–38.

WELLS, J. T. “Enemies Within.” Journal of Accountancy 192 (2001): 31–35.WELLS, J. T. “Control Cash-Register Thievery.” Journal of Accountancy 193 (2002): 88–92.YELLEN, J. L. “Efficiency Wage Models of Unemployment.” American Economic Review 74 (1984):

200–205.ZHANG, Y. “The Effects of Perceived Fairness and Communication on Honesty and Collusion

in a Multi-Agent Setting.” The Accounting Review 83 (2008): 1125–46.

can wages buy honesty? the relationship between relative ... wages buy honesty.pdf · ternative...

Documents